JPH08329254A - Outline extraction device - Google Patents

Abstract

PURPOSE: To realize an automatic editor for picture which doesn't require the intervention of human beings by detecting the intersections of an outline model to split it into plural parts. CONSTITUTION: An outline model deforming part 3 moves outline candidate points stored in an outline model storage part 2 in accordance with a preliminarily determined rule. With respect to the polygonal outline model constituted by connecting the outline candidate points based on a prescribed connection order, an outline model intersection detection part 4 detects whether a certain side on this outline model comes into contact with or crosses another side on the outline model or not. An outline model splitting part 5 splits the outline model into two outline models by two segments first detected by the outline model intersection detection part 4. An extraction completion discrimination part 7 discriminates that outline extraction is completed when all the outline candidate points are not moved and new outline candidate points are not generated neither disappear.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a contour extracting device for extracting a contour of an object captured by using a visual sensor such as a CCD or an infrared camera.

[0002]

2. Description of the Related Art As a technique for extracting the contour of an object from an image picked up by using a CCD camera or the like, information extraction Vol. 30, No. 9, Image processing apparatuses of SNAKES and Japanese Patent Application No. 319051 described in pp.1047 to pp.1057 and the like are known.

According to these techniques, if an outline of the object to be extracted is given in advance, the energy function defined by the outline and the image feature is minimized to obtain an accurate outline of the object. be able to. These techniques can be applied to multimedia devices such as interactive image editing.

[0004]

However, the above-mentioned conventional technique has a problem that a plurality of objects cannot be extracted unless the number of the objects to be extracted is specified in advance.

In SNAKES, it is necessary to give a rough outline of an object as an initial outline in advance. Therefore, when the number of objects and the number of images to be processed are large, the processing load for setting the initial contour increases.

Further, in SNAKES, when it is desired to automatically know how many objects are included in an image, such as detection of the number of intruders in a security system or detection of the number of people in a room in air conditioning control. There was a drawback that could not be dealt with.

In view of the above point, the present invention has an object to provide a contour extracting apparatus capable of extracting contours of a plurality of objects in an image without previously providing the number of objects to be extracted and their outlines. .

The present invention also provides a contour extracting apparatus capable of extracting only the contours of a plurality of objects satisfying a predetermined condition (for example, size) without giving the number of objects to be extracted and their outlines in advance. The purpose is to do.

Another object of the present invention is to provide a contour extracting device capable of extracting a plurality of moving objects without previously giving the number of objects to be extracted and their outlines.

It is another object of the present invention to provide a contour extracting device capable of discriminating between a moving object and a stationary object without giving in advance the number of objects to be extracted and their outlines.

[0011]

According to the present invention for solving the above-mentioned problems, an image storage unit for storing an image and at least one object contained in the image are surrounded, and a contour of the object is extracted. A contour model storage unit that stores a contour model for performing the contour model deformation unit that contracts and deforms the contour model according to a predetermined rule, and a contour model that is contracted and deformed by the contour model deformation unit. When a part touches or intersects with another part of the contour model, the contour model crossing detection unit that detects the touching or crossing, and when the touching or crossing is detected by the contour model crossing detection unit, Based on contact or intersection, a contour model splitting unit that splits the contour model into a plurality, and extraction completion determination that determines whether or not the extraction of the contour of the target object is completed A contour extraction apparatus comprising and.

Further, the contour model storage section may store a plurality of contour candidate points forming the contour model and a connecting order thereof.

Further, the plurality of contour candidate points stored in the contour model storage unit and their connecting order constitute each of a plurality of different contour models obtained by being divided by the contour model dividing unit. A plurality of contour candidate points and their connecting order may be updated.

Further, the contour model deforming unit may move or stop each of the plurality of contour candidate points forming the contour model to shrink and deform the contour model.

Further, the contour extracting device adds at least one new contour candidate point to the contour model and / or deletes at least one contour candidate point in the contour model if a predetermined condition is satisfied. The contour candidate point generation / elimination unit may be further provided.

Further, when each of a plurality of other contour models obtained by dividing the contour model by the contour model dividing unit does not satisfy a predetermined condition, the contour extracting device stores the contour model that does not satisfy the condition in the contour model storage unit. A contour model selection unit to be deleted from may be further provided.

Further, the image storage unit further stores at least one other image which is continuous with the image at a predetermined time interval, and the contour extracting device is stored in the image storage unit. Based on multiple consecutive images
A motion detection unit that detects a movement vector may be further provided.

Further, the image storage unit stores a plurality of continuous images at a predetermined time interval, and the contour extracting device selects the contour extracting device for at least one of the plurality of images. It is also possible to further include a motion detection unit for each contour model that detects a movement vector within a predetermined range of each contour candidate point that constitutes the contour model of at least one object extracted by.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. (Embodiment 1) FIG. 1 shows a first embodiment of a contour extracting apparatus of the present invention.
It is a block diagram which shows the Example of. In FIG. 1, 1 is an image storage unit that stores an image of a scene including an object,
Reference numeral 2 denotes a contour model storage unit that stores contour candidate points for searching contour positions of an object in an image stored in the image storage unit 1, and reference numeral 3 denotes contour contour points stored in the contour model storage unit 2 in advance. The contour model transforming unit 4 which moves according to a predetermined rule is a contour model crossing for detecting the crossing in a contour model configured by connecting contour candidate points according to a predetermined rule. A detection unit 5 divides a contour model composed of contour candidate points into a plurality of contour model divisions when an intersection is detected by the contour model intersection detection unit 4, and 6 an intersection is detected by the contour model intersection detection unit 4. If not, a contour candidate point generation / disappearance unit for newly adding or erasing contour candidate points, and 7 is an extraction completion determining whether or not the contour extraction of the object in the image is completed. A determination unit.

By the way, 8 is connected to each part of 1 to 7,
The control unit controls the start and stop of the operation of each unit or the exchange of data between each unit. The control unit 8 is a computer including a memory such as a ROM or a RAM and a CPU, and an imaging device 9 such as a CCD camera or a scanner as shown in FIG. 1 and a CRT through various interfaces. And a display device 10 and an input device 11 such as a keyboard and a mouse. The image storage unit 1 and the contour model storage unit 2 are normally RAMs.
Composed of.

Next, the operation of this embodiment will be described.
FIG. 2 is a flow chart showing the operation of the contour extracting apparatus of this embodiment. FIG. 3 is a diagram showing an example of extracting the contour of a cell from a micrograph in order to make the operation of the present embodiment easier to understand. This means that the present invention can be applied to automated systems such as medical diagnosis of cancer cells. The image storage unit 1 stores an image of a cell micrograph captured by an imaging device 9 such as a scanner and digitized. Each pixel data of the digitized image is stored, for example, by representing its brightness with 8 bits. The operation procedure of this embodiment is shown below. [STEP1a] Initial placement of contour candidate points Configure one polygon (contour model) that includes all the objects included in the image stored in the image storage unit 1
The n points are initial contour candidate points v _i (x _i (0), y _i (0)) (i = 1,2,
, N) are stored in the contour model storage unit 2. i
Shows the order of connecting the contour candidate points for constructing the contour model by connecting the contour candidate points. In addition, the operator is
The initial contour candidate points may be stored in the contour model storage unit 2 by operating the input device 11 such as a keyboard or a mouse while looking at the image displayed at 0. Alternatively,
A point obtained by equally dividing a rectangular side substantially equal to the size of the image into n may be stored in advance in the contour model storage unit 2 as an initial contour candidate point. [STEP2a] Deformation of contour model The contour model transformation unit 3 sets the positions of all contour candidate points to 1
Move it twice. As a result, the contour model in which the contour candidate points are connected is deformed. Contour model transformation unit 3
Is the energy E _snake (v _i ) of the predefined contour model
Each contour candidate point is moved in the direction that minimizes.

For example, the contour model transformation unit 3 uses the contour candidate
Point v_i(x_i, y_i) (i = 1,2, ..., n), E _snake(v_i) Steepest descent
Lower vector (-∂E_snake/ ∂x_i, -∂E_snake/ ∂y_i)
Then, using (Equation 1) of the following equation,
Move each contour candidate point in parallel independently for each point.
It

[0023]

[Equation 1] x _i (t + 1) = x _i (t) -Kx (∂E _snake / ∂x _i ) y _i (t + 1) = y _i (t) -Ky (∂E _snake / ∂y _i ) Note that t (i = 0, 1, ...) In (Equation 1) represents the number of movements of each contour candidate point. Kx> 0 and Ky> 0 are constants that control the movement amount of the contour candidate points.

However, when the energy E _snake (v _i ) increases due to the movement, the contour model deforming section 3 does not move the contour candidate points.

Further, the contour model transforming section 3 uses the flag F
(v_i) (i = 1,2, ..., n)
It is stored in the memory 2. That is, the contour model transformation unit 3 moves
F (v_i) = 1, do not move
F (v _i) = 0.

By the way, the energy E _{snake of the} contour model
(v _i) may be, for example, is the sum of each of the energy terms represented by the following from equation (2) (6). (1) E _spline that represents the smoothness of the contour model

[0027]

[Equation 2]

When each contour candidate point is moved so as to minimize (Equation 2), the contour model contracts. (2) E _area corresponding to the area of the closed area enclosed by the contour model

[0029]

(Equation 3)

Note that x _{n + 1} = x ₁ and y _{n + 1} = y ₁ .

When each contour candidate point is moved so as to minimize (Equation 3), each contour candidate point moves in a direction perpendicular to the contour model. If E _area is used, the contour of the concave object can be extracted. (3) E _dist that averages the distances between contour candidate points

[0032]

[Equation 4]

Note that d _av represents the average distance between contour candidate points. 4) Gradient magnitude E _{edge of} image brightness I (v _i )

[0034]

(Equation 5)

When each contour candidate point is moved so as to minimize (Equation 5), each contour candidate point moves in the direction in which the brightness gradient increases, that is, in the direction toward the contour portion of the object. (5) Image brightness value E _intens

[0036]

(Equation 6)

When each contour candidate point is moved so as to minimize ( _Equation 6), when w _intens > 0, each of the contour candidate points becomes w _intens <0 in the direction in which the brightness becomes smaller (darker). In the case of, it moves in the direction in which the brightness becomes larger (brighter).

In addition, in (Equation 2) to (Equation 6)
Each of w _sp1 , w _sp2 , w _area , w _dist , w _edge , and w _intens is a weighting coefficient of the energy term.

If the contour model is deformed so as to minimize the energy E _{s nake} (v _i ) of the contour model, which is the sum of the energy terms of ( _Equation 2) to ( _Equation 6), the contour model contracts. However, in the end, the vector by each energy term of (Equation 2) to (Equation 6) that moves each contour candidate point is
It stops at the contour of the target object. The contour model formed by the contour candidate points at this time is extracted as the contour of the object. [STEP3a] Detecting Contour of Contour Model For a contour model forming a polygon formed by concatenating each contour candidate point based on a predetermined concatenation order, the contour model intersection detection unit 4 detects that one side on the contour model , It detects whether or not it is in contact with or intersects with another side on the contour model. However, two adjacent sides are excluded from the detection target. If each contour candidate point is moved in pixel units, in most cases, it is detected when one of the end points of the side comes into contact with another side (see FIG. 4).

For example, the contour model intersection detector 4 may detect contact or intersection as follows.

Assuming that the line segment v _i v _{i + 1} and the line segment v _j v _{j + 1} on the contour model have an intersection, a real number p satisfying the following (Equation 7)
There are (0 ≦ p ≦ 1) and q (0 ≦ q ≦ 1).

[0042]

[Equation 7] _{p (v i + 1 -v i} ) + v i = q (v j + 1 -v j) + v j ( number 7) is a system of equations p, on q, the following (number 8 ) Has a solution when the determinant det calculated by is not 0.

[0043]

(8) det = (x _i -x _{i + 1} ) (y _{j + 1} -y _j )-(x _{j + 1} -x _j )
(y _i -y _{i + 1} ) When det = 0, the line segment v _i v _{i + 1} and the line segment v _j v _{j + 1} may match. A match should be detected.

The following (Formula 9) is the line segment v _i v _{i + 1} and the line segment v _j v _{j + 1.}
Represents the position of the point of contact or intersection when touches or crosses.

[0045]

(9) p = ((y _{j + 1} -y _j ) (x _i -x _j ) + (x _j -x _{j + 1} ) (y _i -y _j )) / det q = ((y _{i + 1} -y _i ) (x _i -x _j ) + (x _i -x _{i + 1} ) (y _i -y _j )) / det In short, the contour model intersection detection unit 4 sequentially changes i and j. Then, when det ≠ 0, when p and q calculated by (Equation 9) satisfy 0 ≦ p ≦ 1 and 0 ≦ q ≦ 1, respectively, the line segment
The v _i v _{i + 1} and the line segment v _j v _{j + 1} are regarded as touching or intersecting at one point (p, q) on the contour model, and the touching or crossing is detected. The contour model intersection detection unit 4 terminates the detection operation when the contact or the intersection is first detected.

The control unit 8 executes STEP4a when contact or intersection is detected, and otherwise STEP5a.
To execute. [STEP4a] Discontinuity of contour model Fig. 4 shows an example of a typical intersection that occurs when E _area is used as the energy of the contour model. The contour model dividing unit 5 divides the contour model into a plurality of pieces. When the contour model dividing unit 5 completes the division process, the control unit 8 executes STEP2a again.

The contour model dividing unit 5 uses the line segment v _i v _{i + 1} and the line segment v _j v _{j + 1} having the intersections first detected by the contour model intersection detection unit 4 to calculate the contour model (= FIG. 4). Contour model 1 shown in
And 2) are divided into contour model 1 and contour model 2. That is, the contour model division unit 5 uses the line segment v _i v
_A contour model point v _i having a previous number in _{i + 1} and a contour candidate point v _{j + 1} having a subsequent number in the other line segment v _j v _{j + 1} having an intersection are connected to each other to form a contour model ( = Split contour models 1 and 2) into contour model 2. Similarly, the contour model division unit 5 determines the contour candidate point v _j having the previous number in the line segment v _j v _{j + 1} having the intersection point and the contour candidate point v _{j in the} other line segment v _i v _{i + 1} having the intersection point. The contour model (= contour model 1 and 2) is divided into contour models 1 by connecting the contour candidate points v _{i + 1} having numbers.

As a result, the contour models (= contour models 1 and 2) are connected in the order of {v ₁ , ..., v _i , v _{j +1} , ..., v _n } and {v _{i + 1} ,.
,, v _j } is divided into a set of contour candidate points.

After the above division, the control unit 8 determines whether contact or intersection has occurred also with respect to the line segment after the line segment v _i v _{i + 1 for} which the contact or intersection detection processing has not been performed in STEP 3a. The contour model intersection detection unit 4 is caused to detect whether or not it is. Also in this case,
In the line segment after the line segment v _i v _{i + 1} , the detection process is interrupted at the first detected line segment, and the division process of STEP 4a is applied again. Such recursive processing is repeated until all line segments on the contour model (= contour models 1 and 2) are completed. [STEP5a] Generation / disappearance determination of contour candidate points The contour candidate point generation / disappearance unit 6 performs generation / disappearance of new contour candidate points.

The contour candidate point generation / elimination unit 6 newly generates contour candidate points in the following cases, for example. The contour candidate point generation / annihilation unit 6 determines that the distance between adjacent contour candidate points is | v.
_{When i + 1-} v _i |> D _TH is satisfied, a new contour candidate point is generated between two points v _i and v _{i + 1} . However, D _TH is a predetermined maximum distance between contour candidate points. At this time, the contour candidate point generation / annihilation unit 6 sets G _flag = 1 if the contour candidate points are generated, and sets G _flag = 0 if they are not generated.

Further, the contour candidate point generation / elimination unit 6 eliminates the contour candidate points in the following cases, for example. The contour candidate point generation / annihilation unit 6 uses cos θ> θ _TH (θ is v _i v _i-1 and v _i v
_The contour candidate points v _i satisfying the ( _{i + 1} angle) are eliminated. However, θ _TH is a threshold value for determining the sharpness of a predetermined contour model. At this time, the contour candidate point generation / annihilation unit 6 sets D _flag = 1 if the contour candidate points are to be eliminated, and sets D _flag = 0 if they are not eliminated.

The additional generation of contour candidate points compensates for the number of contour candidate points per object, which decreases due to division,
The contours of multiple objects can be extracted correctly.

Further, according to this disappearance procedure, the contour candidate points stopped at positions other than the contour of the object are deleted, and erroneous contour extraction can be avoided. [STEP6a] Extraction completion determination The extraction completion determination unit 7 does not move to all contour candidate points (∀i;
When F (v _i ) = 0) and there is no generation / disappearance of new contour candidate points (G _flag = 0 and D _flag = 0), it is determined that the contour extraction is completed.

When the extraction completion judgment unit 7 judges that the contour extraction is completed, the control unit 8 terminates the contour extraction process, and otherwise returns to STEP 2a.

According to the above procedure, as shown in FIG.
An initial contour model (Fig. 5 (a)) that includes the entire set of objects (cells) is sequentially crossed and divided in the transformation process (Figs. 5 (b) to (e)), so that a plurality of objects can be obtained. It can be seen that each can be extracted independently (Fig. 5 (f)).

As described above, according to this embodiment,
By detecting the intersection of the contour models and dividing the intersection into a plurality of contours, the contours of a plurality of objects in the image can be automatically extracted. (Embodiment 2) Next, a second embodiment of the contour extracting apparatus of the present invention will be described. In the first embodiment, it has been explained that the contours of a plurality of objects in the image can be automatically extracted by detecting the intersection or contact of the contour models and dividing the contour model into a plurality of pieces. The present embodiment takes into consideration that only objects that satisfy the specified conditions can be selectively extracted. For example, the condition to be specified is the size of the object. In the first embodiment, cell contour extraction has been described as an example, but in medical diagnosis, it is often necessary to extract only cells of a specific size.

FIG. 6 is a block diagram of this embodiment. Since the same components as those of the first embodiment shown in FIG. 1 are designated by the same reference numerals, detailed description of the components designated by the same reference numerals is omitted. On the basis of the configuration of the first embodiment, this embodiment further uses the contour model stored in the contour model storage unit 2 when the contour model does not satisfy a predetermined condition. A contour model selection unit 12 to be deleted from the contour model storage unit 2 is provided. As a result, it is possible to selectively extract the target object in consideration of the size and the like.

Next, the operation of this embodiment will be described.
FIG. 7 is a flowchart showing the operation of the contour extracting device in this embodiment. STEP 1b in the flowchart
And STEP3b to STEP7b are exactly the same as STEP1a and STEP2a to STEP6a in the first embodiment, respectively. Therefore, different STEP2b will be described. [STEP2b] Selection of Contour Model The contour model selection unit 12 deletes a contour model that does not satisfy a predetermined condition from the contour model storage unit 2.

For example, the threshold value for E _area in (Equation 3) is set as the predetermined condition. And
If the E _{area of} a contour model is below its threshold,
The contour model is deleted from the contour model storage unit 2.

By the above procedure, as shown in FIG. 8, by sequentially erasing the contour model having no predetermined area value, it is possible to extract only relatively large cells in the central portion of the image.

As described above, according to this embodiment,
By calculating the area (E _area ) of the closed region by each contour model at the time when the contour extraction of each object is completed or when the contour extraction of all the objects is completed, Contours that are smaller, larger than a predetermined area, within a predetermined area, or outside a predetermined area can be deleted. As a result, it is possible to extract only the target object having a desired size. (Embodiment 3) Next, a third embodiment of the contour extracting apparatus of the present invention will be described. In the first and second embodiments, the extraction of stationary plural objects has been described. The present embodiment considers extraction of a plurality of moving objects.
Basically, a moving object is extracted by detecting the presence or absence of movement of an edge and smoothly connecting only the moving edge by a contour model.

Specifically, the movement vector is obtained by using a plurality of images which are taken continuously in terms of time. When the contour model is deformed, when the magnitude of the movement vector at the position of each contour candidate point of the contour model is equal to or lower than the movement speed of the extraction target (for example, 1 pixel / frame), the image energy (for example, E _edge ) The weighting factor (eg W _edge ) of 0
And As a result, the contour model can be drawn only to a moving edge, and the moving edges can be smoothly connected to extract the contours of a plurality of moving objects.

FIG. 9 is a block diagram of this embodiment. Since the same components as those in the first embodiment shown in FIG. 1 are designated by the same reference numerals, detailed description of the components having the same reference numerals will be omitted. Based on the configuration of the first embodiment, this embodiment
Furthermore, it is provided with a motion detection unit 13 that detects motion from a plurality of images that are temporally continuously captured. Further, the image storage unit 1 stores a plurality of images that are continuously captured.

Next, the operation of this embodiment will be described.
FIG. 10 is a flow chart showing the operation of the contour extracting apparatus of this embodiment. STEP2c and STEP4c to STEP7c in the flowchart are respectively STEP1a of the first embodiment.
And since it is the same as STEP3a ~ STEP6a, different STEP1c,
Only STEP3c will be explained. In this embodiment, a person walking in a living room as shown in FIG. 11 is extracted as a moving object. [STEP1c] Motion Detection The motion detection unit 13 detects a movement vector from a plurality of images stored in the image storage unit 1.

For example, the motion detection unit 13 obtains a movement vector by the least square method based on the gradient method. In the gradient method, the partial derivative of the image brightness I (x, y) with respect to time and space,
The following constraint on the apparent velocity (u, v) = (dx / dt, dy / dt) of the object is used.

[0066]

[Expression 10] Ix (x, y) u + Iy (x, y) v + It (x, y) = 0 Note that Ix = ∂I / ∂x, Iy = ∂I / ∂y, It = ∂I / T.

Assuming that the movement vector is the same in the local small area in the image, that is, the constraint equation of the equation (10) has the same solution in the local small area S, the movement vector is The following (Equation 11) is obtained as (u, v) that minimizes.

[0068]

[Equation 11] E = Σ (Ix (i, j) u + Iy (i, j) v + It (i, j)) ² (i, j) εS Therefore, by the following (Equation 12), ∂E / ∂u = 0, ∂E / ∂v
It is sufficient to find (u, v) that satisfies = 0.

[0069]

(12) u = (ΣIxIy · ΣItIy-Σ (Iy) ² · ΣItIx) / Det v = (ΣIxIy · ΣItIx-Σ (Ix) ² · ΣItIy) / Det However, Ix, Iy, It are respectively ( Ix shown in equation 11)
(i, j), Iy (i, j), It (i, j), where i and j are the x and y coordinates of the pixels belonging to the small area S in the image, and Σ is (i, j
j) Calculate for εS. Det is calculated by the following formula.

[0070]

[Formula 13] Det = Σ (Iy) ² · Σ (Ix) ^2- (ΣIxIy) ² [STEP3c] Deformation of contour model The contour model transformation unit 3 sets the positions of all contour candidate points to 1
Move it twice. At this time, the weight coefficient w _edge of E _edge is set to 0 at the position where the motion is not detected in STEP 1c. As a result, the contour model is drawn only to the edge in which the motion is detected.

According to the above procedure, as shown in FIG. 12, the motion is detected by the gradient method (FIG. 12 (a)), and when the contour candidate points are moved, the E _edge is detected at the position where the motion is not detected.
If the weighting coefficient w _{edge of is set} to 0, the contour model can be attracted only to the moving edge. As a result, the contour extracting device according to the present embodiment can extract, for example, a pedestrian as the plurality of moving objects (FIG. 12).
(b)).

As described above, according to this embodiment,
It is possible to extract a plurality of moving objects by detecting a motion from a plurality of continuously captured images and smoothly connecting only the moving edges. (Fourth Embodiment) Next, a fourth embodiment of the contour extracting apparatus of the present invention will be described. In the first to third embodiments, the CCD
The device for extracting an object from a visible image obtained from a camera or the like has been described. However, it is often difficult to extract a human body or the like only with information on a visible image. In this embodiment, a heat generating object such as a human body is extracted by using a thermal image obtained from an infrared camera, and at the same time, a stationary heat generating device such as a CRT of a personal computer can be distinguished from the human body.

FIG. 13 is a block diagram of this embodiment. FIG.
The same components as those of the first embodiment shown in are attached with the same numbers, and detailed description of the same components is omitted. On the basis of the configuration of the first embodiment, the present embodiment further includes a contour model-specific motion detection unit 14 that detects motion in the vicinity of the contour candidate points stored in the contour model storage unit 2. ing. In addition, the image storage unit 1 stores a plurality of thermal images captured continuously in time.

As a result, the contour model in which the motion is detected can be distinguished from the contour model for the stationary heat-generating device, assuming that it is the contour model for the human body.

Next, the operation of this embodiment will be described.
FIG. 14 is a flow chart showing the operation of the contour extracting apparatus of this embodiment. STEP 1d in the flow chart
Steps 6d are exactly the same as steps 1a to 6a of the first embodiment, so only different steps 7d will be described. [STEP7d] Detection of Motion by Contour Model The motion detection unit by contour model 14 detects motion in the vicinity of the contour candidate points forming at least one contour model determined to be extracted in [STEP6d].

For example, the motion may be detected by using the gradient method in the third embodiment. When motion is detected for a predetermined number of contour candidate points of the contour model, it is determined that the contour model is extracting a moving heat-generating object (for example, a human body).

According to the above procedure, as shown in FIG. 15, the human body and the stationary heat generating device can be discriminated by detecting the movement for each contour model.

As described above, according to the fourth embodiment, the human body and the stationary heat-generating device are distinguished by detecting the movement for each contour model using the thermal images continuously obtained in time. Is possible.

[0079]

As is apparent from the above, according to the present invention, the contours of a plurality of objects in an image are automatically extracted by detecting the intersection of contour models and dividing them into a plurality of divisions. You can Therefore, it is possible to realize an automatic image editing device that does not require human intervention.

Further, according to the present invention, by deleting the contour model which does not satisfy the predetermined condition, for example, only the object having the predetermined size can be extracted. Therefore, it can be applied to, for example, an automated system for medical diagnosis that targets only cells of a specific size.

Further, according to the present invention, it is possible to extract a plurality of moving objects by detecting a motion from a plurality of continuously picked-up images and smoothly connecting only moving edges. Therefore, for example, since only the person walking in the living room can be extracted, the present invention can be applied to the air conditioning control adapted to the human activity state.

Further, according to the present invention, the human body and the stationary heat generating device can be distinguished by detecting the movement for each contour model using a plurality of consecutively taken thermal images. Therefore, it is possible to realize detailed air conditioning control according to the situation of the person in the room, such as opening the air outlet of the air conditioner in a place where the human body is present.

[Brief description of drawings]

FIG. 1 is a block diagram of a first embodiment according to a contour extraction device of the present invention.

FIG. 2 is a flowchart showing an operation procedure of this embodiment.

FIG. 3 is a diagram showing an example of an image including an object (cell).

FIG. 4 is a diagram showing an example of intersection / division of a contour model.

FIG. 5 is a diagram showing how an object (cell) is extracted.

FIG. 6 is a block diagram of a second embodiment according to the contour extraction device of the present invention.

FIG. 7 is a flowchart showing an operation procedure of this embodiment.

FIG. 8 is a diagram showing how an object (cell) is extracted in consideration of size.

FIG. 9 is a block diagram of a third embodiment according to the contour extraction device of the present invention.

FIG. 10 is a flowchart showing an operation procedure of this embodiment.

FIG. 11 is a diagram showing an example of an image including a moving object (pedestrian).

FIG. 12 is a diagram showing a motion detection result and a moving object (pedestrian) extraction result.

FIG. 13 is a block diagram of a fourth embodiment according to the contour extraction device of the present invention.

FIG. 14 is a flowchart showing an operation procedure of this embodiment.

FIG. 15 is a diagram showing a result of identifying a human body and a heat-generating device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image storage unit 2 Contour model storage unit 3 Contour model transformation unit 4 Contour model intersection detection unit 5 Contour model division unit 6 Contour candidate point generation / disappearance unit 7 Extraction completion determination unit 8 Control unit 9 Imaging device 10 Display device 11 Input device 12 Contour Model Selection Unit 13 Motion Detection Unit 14 Contour Model-Specific Motion Detection Unit

Claims (8)

[Claims] 1. An image storage unit that stores an image; a contour model storage unit that surrounds at least one object included in the image and stores a contour model for extracting the contour of the object; In accordance with the rule, in the contour model deforming unit that contracts and deforms the contour model, and in the contour model that is contracted and deformed by the contour model deforming unit, a part of the contour model touches or intersects with another part of the contour model. If a contact or intersection is detected by the contour model intersection detection unit that detects the contact or intersection, and if the contact or intersection is detected by the contour model intersection detection unit, the contour that divides the contour model into a plurality of sections based on the contact or intersection. A contour extraction device comprising: a model division unit; and an extraction completion determination unit that determines whether or not the contour extraction of the object is completed. 2. The contour extraction device according to claim 1, wherein the contour model storage unit stores a plurality of contour candidate points forming the contour model and a connecting order thereof. 3. A plurality of contour candidate points stored in the contour model storage unit and their connecting order constitute each of a plurality of different contour models obtained by being divided by the contour model dividing unit. The contour extracting apparatus according to claim 2, wherein the contour extracting points are updated to a plurality of contour candidate points and their connecting order. 4. The contour model deforming unit moves or stops each of a plurality of contour candidate points forming the contour model to shrink and deform the contour model. The described contour extraction device. 5. If at least one new contour candidate point is added to the contour model if a predetermined condition is satisfied, and / or
Alternatively, the contour extracting device according to claim 2, further comprising a contour candidate point generating / erasing unit that deletes at least one contour candidate point in the contour model. 6. A contour model selection for deleting the unsatisfactory contour model from the contour model storage unit when each of a plurality of different contour models obtained by being divided by the contour model dividing unit does not satisfy a predetermined condition. The contour extracting device according to claim 2, further comprising a section. 7. The image storage unit further stores at least one other image continuous with a predetermined time interval with respect to the image, and the contour extraction device is stored in the image storage unit. The contour extraction device according to claim 1, further comprising a motion detection unit that detects a movement vector based on a plurality of continuous images that are present. 8. The image storage unit stores a plurality of continuous images at a predetermined time interval, and the contour extracting device selects a contour extracting device for at least one image of the plurality of images. 3. The motion detection unit for each contour model for detecting a movement vector within a predetermined range of each contour candidate point constituting the contour model of at least one target object extracted by 1.
7. The contour extraction device according to any one of 1 to 6.